Back to EveryPatent.com
| United States Patent |
5,027,287
|
|
Artigalas
, et al.
|
June 25, 1991
|
Device for the digital processing of images to obtain special
geometrical effects
Abstract
The obtaining of a geometrical effect, with an image, is done by a
two-dimensional filter with variable characteristics controlled by a
computer to compute the points of a pre-established conversion grid,
superimposed on the image, as a function of neighboring points of the
original image and of the desired effect. A memory contains the original
image. Auxiliary images, obtained by successive reductions, in fixed
proportion, of the original image, are stored and a point of the
conversion grid is computed by selecting that of the memories for which
the reduction factor enables the easiest computation while, at the same
time, preserving the quality of the image.
| Inventors:
|
Artigalas; Max (Le Chesnay, FR);
Grimaldi; Jean-Luc (Ermont, FR)
|
| Assignee:
|
Thomson Video Equipement (Cergy Saint Christophe, FR)
|
| Appl. No.:
|
362349 |
| Filed:
|
June 6, 1989 |
Foreign Application Priority Data
| Current U.S. Class: |
345/670; 348/580; 382/298 |
| Intern'l Class: |
G06F 015/66; H04N 009/74 |
| Field of Search: |
364/521,518
382/47,44,56
358/22
|
References Cited
U.S. Patent Documents
| 4432009 | Feb., 1984 | Reitmeier et al. | 358/22.
|
| 4665433 | May., 1987 | Hinson et al. | 358/22.
|
| 4760605 | Jul., 1988 | David et al. | 382/47.
|
| 4774678 | Sep., 1988 | David et al. | 364/518.
|
| 4831447 | May., 1989 | Lake, Jr. | 358/183.
|
| Foreign Patent Documents |
| 2158322 | Nov., 1985 | GB.
| |
| 2174861 | Nov., 1986 | GB.
| |
Other References
Toute L'Electronique, No. 460, Jan. 1981, pp. 23-28, A. Le Toquin.
Computer Graphics & Image Processing, vol. 13, No. 2, Jun. 1980, pp.
127-141, C. Braccini et al.
|
Primary Examiner: Harkcom; Gary V.
Assistant Examiner: Bayerl; Raymond J.
Attorney, Agent or Firm: Oblon, Spivak, McClelland, Maier & Neustadt
Claims
What is claimed is:
1. A digital processing device for obtaining special geometrical effects
from an original image formed by digital signals representing points
distributed along the intersections of a rectangular grid formed by rows
and columns, said device comprising an image memory (MO) to preserve the
original image, a two-dimensional filter (F) with variable characteristics
and a computer to control, as a function of a pre-established conversion
grid, and as a function of the compression factors of this conversion
grid, the computation, by the filter, of the digital values to be assigned
to the intersections of the conversion grid, said device also comprising
n, where n is a whole number at least equal to 1, auxiliary
two-dimensional filters, with fixed characteristics, placed in series and
each having one output, where the first auxiliary filter in the series
receives the digital signals of the original image, n memories (M1-M3)
respectively connected to the outputs of the n auxiliary filters and a
change-over switch 2 to switch one of the n+1 memories (M0-M3) over to the
filter with variable characteristics F, with the computer activating the
change-over switch as a function of the compression factors G of the
conversion grid.
2. A device according to claim 1, wherein the auxiliary filters all have
the same compression rate.
3. A device according to claim 2, wherein the compression rate of the
auxiliary filters is 2.
Description
BACKGROUND OF THE INVENTION
1. Field of the Invention
The present invention concerns the digital processing of images, for the
obtaining of special geometrical effects. The processing is done using an
original image formed by digital signals, representing points distributed
along the intersections of a rectangular grid formed by rows and columns.
The processing consists in modifying the geometry of the original image to
obtain, for example, the appearance of a projection of the image on a
curved surface, or a reduction, namely a simple change in scale of the
image.
In a known method to achieve these special effects, the conversion grid,
the rows and columns of which can be curved, is determined, this grid is
superimposed on the original rectangular grid, and their points of
intersection are determined by interpolation, using the neighboring points
on the original grid; the points of the conversion grid are then carried
over to a rectangular grid, having the same number of rows and columns and
the special effect is obtained. To correct aliasing phenomena, namely the
conversion of high frequencies of the spectrum into low frequencies, a
known method lies in determining each point of intersection by bringing
into play two factors, called local compression factors, equal to the
respective norms of the vectors respectively tangential to the row and
column at the point considered.
It must be noted that the smaller the final image as compared with the
original image, the greater is the number of neighboring points that are
to be taken into account in determining, by interpolation, the points of
the new grid in order to minimize the aliasing phenomena. Thus, when the
final image is smaller than the original image, the prior art devices
require long computation periods to prevent the aliasing phenomenon.
SUMMARY OF THE INVENTION
The present invention is aimed at preventing or, at least, reducing these
drawbacks. This is obtained by successively reducing, and storing, the
image, and then obtaining the desired image from the images stored. The
stored image is selected, for the determining of each point of the final
image, as a function of the local compression factors of the point
considered and of the coefficients of reduction of the stored images.
According to the invention, there is provided a digital processing device
for obtaining special geometrical effects from an original image formed by
digital signals representing points distributed along the intersections of
a rectangular grid formed by rows and columns, said device comprising an
image memory (M0) to preserve the original image, a two-dimensional filter
(F) with variable characteristics and a computer to control, as a function
of a pre-established conversion grid, and as a function of the compression
factors of this conversion grid, the computation, by the filter, of the
digital values to be assigned to the intersections of the conversion grid,
said device also comprising n, where n is a whole number at least equal to
1, auxiliary two-dimensional filters (F1-F3), with fixed characteristics,
placed in series and each having one output, where the first auxiliary
filter in the series receives the digital signals of the original image, n
memories (M1-M3) respectively connected to the outputs of the n auxiliary
filters and a change-over switch 2 to switch one of the n+1 memories
(M0-M3) over to the filter with variable characteristics F, with the
computer activating the change-over switch as a function of the
compression factors G of the conversion grid.
BRIEF DESCRIPTION OF THE DRAWINGS
The present invention will be better understood and other characteristics
will appear from the following description, and the figures pertaining
thereto, of which:
FIG. 1 shows two grids used to produce a geometrical effect in an image;
FIG. 2 is a diagram of a prior art device for the obtaining of a
geometrical effect;
FIG. 3 is a diagram of a device according to the invention for the
obtaining of a geometrical effect;
FIG. 4 is a flowchart relating to FIG. 3.
In the various diagrams, the precise synchronization devices, coming under
standard technology, have not been shown in order to make the diagrams
clearer and to simplify the explanation.
MORE DETAILED DESCRIPTION
FIG. 1 gives a partial view of two grids, one rectangular, formed by
horizontal rows X and vertical columns Y, while the other is formed by
curved rows and columns U, V. The rectangular grid represents, by its
intersection points, the position of the points of an original image, and
the points of the original image are known by digital signals. This grid
shall hereinafter be called an original grid. The other grid is a
transposition grid, namely a grid for which the intersection points,
computed by interpolation from the neighbouring points of the original
grid, are carried over to the intersections of a rectangular grid having
the same number of rows and columns as the transposition grid. It is
through this process that the achieving of special geometrical effects
from an original image is known.
It should be further noted that, following Shannon's theorem, it is
necessary to determine appropriate sampling spectra in the directions of
the curves U and V, to prevent aliasing phenomena that create faults. For
this purpose, the associating, with each point (u, v), of two factors
called local compression factors, along the tangents to the curves U and
V, is known. These local compression factors are equal to the respective
norms of the vectors tangential to the curves U and V:
C.sub.U =.sqroot.(dU/dx).sup.2 +(dU/dy).sup.2
C.sub.V =.sqroot.(dV/dx).sup.2 +(dV/dy).sup.2
In these formulae, C.sub.U and C.sub.V represent, respectively, the
compression factors along U and V. Values that are a little less precise,
but simpler to obtain, can be determined by means of the chord that
subtends two consecutive points. Thus at the point (u2, v2) of FIG. 1
C.sub.U =.sqroot.(x2,2-x1,2).sup.2 +(y2,2-y2,1).sup.2
C.sub.V =.sqroot.(x2,2-x2,1).sup.2 +(y2,2-y2,1).sup.2
These compression factors, irrespectively of how they are defined, depend
on the curves U and V, which define the special effect, and on the pitch
of the original grid. The greater the compression factor, the greater is
the number of points of the original grid to be taken into account to
determine, by interpolation, a point of the conversion grid in order to
minimize the aliasing phenomenon. FIGS. 3 and 4 will make it possible to
show how, in order to avoid making lengthy computations or having a large
number of electronic circuits to take into account a large number of
points of the original grid when the compression factors are great, it is
possible to do the computations not from the original grid alone, but from
a set of grids deduced from one another by simple reductions using
two-dimensional filters placed in series.
FIG. 2 is a diagram illustrating the prior art, referred to above, to
achieve a special geometrical effect. The digital signals Sn, representing
points of an original image distributed along the intersection of a
rectangular grid are stored in a memory M.sub.O called an original image
memory. A computer 1 has the data of the conversion grid: the address of
the points of the original grid and compression factors to be taken into
account for each intersection point of the conversion grid. The computer 1
activates, by an addressing signal A, the dispatch from the memory
M.sub.O, to a two-dimensional filter F, of the signals of the original
image to be used for the computation, by interpolation, of a point of the
conversion grid. The computer also sends the filter F a signal K giving
the two compression factors relating to the point to be computed. The
filter F is a two-dimensional filter with a variable compression rate
controlled by the signal K. This filter gives, either directly if the work
is done in real time, or with storage in a memory grid Ms, a signal St
corresponding to the signal Sn converted according to the special
geometrical effect desired.
Should the special effect desired be a simple reduction, i.e. a change in
scale of the image, the computation by interpolation and the circuit of
FIG. 2 are simplified: the computer 1 is eliminated and a two-dimensional
filter with a fixed compression rate is used as the filter F, and the
computation by interpolation is then done systematically in the same way
for all the points of the conversion grid.
FIG. 3 shows a circuit with which the computations, done by interpolation,
of the points of the conversion grid are simplified in comparison with the
corresponding computations done in the circuit of FIG. 2.
In the circuit according to FIG. 3, the signal Sn is always sent into an
original image memory M.sub.O, but it is also sent through three auxiliary
filters F.sub.1, F.sub.2, F.sub.3 to three image memories M.sub.1,
M.sub.2, M.sub.3. The auxiliary filters F.sub.1, F.sub.2, F.sub.3 are
two-dimensional filters with fixed compression rate, having a value 2 in
the example described. It is therefore a simple image reduction which, as
indicated during the description of FIG. 2, is done very simply.
Thus, in the memories M.sub.0, M.sub.1, M.sub.2, images are stored with an
increasingly low resolution since their compression factor 2.sup.N, with
respect to the original image, is respectively 2.sup.0 =1, 2.sup.1 =2,
2.sup.2 =4, 2.sup.3 =8.
A computer 1 having, for one and the same special effect, the data
corresponding to the same conversion grid as the computer of FIG. 2, will
thus be capable of choosing between the contents of the memories M.sub.0
to M.sub.3 to trigger computation, by interpolation, of a point of the
conversion grid. This computation is done with the same two-dimensional
filter F as in the case of FIG. 2. The filter F receives the image signals
stored in the memories M.sub.0 to M.sub.3 through a change-over switch 2,
with four inputs and one output.
The choice of one of the four memories M.sub.0 to M.sub.3 is done as a
function of the smallest of the two compression factors associated with
the point to be determined. If 2.sup.a and 2.sup.b are these factors, with
a smaller than b, the memory chosen is that for which the compression
factor 2.sup.N is such that:
N.ltoreq.a <N +1
it being understood that, should a-1 be greater than the maximum value of
N, i.e. greater than 3, it is the memory for which the compression rate
equals 2.sup.3 =8 that is chosen. A correction is made to the choice if
the value of a should be smaller, by more than 10%, than the values N=Ng
of the compression factor 2.sup.N of the memory chosen to compute the
preceding point. This correction, which is not indispensable, consists in
retaining the preceding value Ng. This makes it possible to reduce the
faults introduced by the switching over of the memories. The method of
choosing among the four memories, that has just been described, is
illustrated by the flowchart of FIG. 4: the switching over of the memory
by the change-over switch 2 is permitted only if a is outside the range
Ng-10%, inclusive of the limit, or Ng+1 excluding the limit.
The computer 1 gives a signal G to activate the change-over switch 2. This
signal G is obtained according to the flowchart of FIG. 4. It should be
noted, besides, that the address signals, A, sent by the computer 1 to the
memory M.sub.0 to M.sub.3, are the same as in the case of FIG. 2 and are
the same irrespectively of the memory chosen. In effect, in the example
described, where the memories M.sub.0 to M.sub.3 contain a number of
signals that decrease according to a geometrical sequence of 2, the signal
G is formed by groups of eight addresses concerning eight points of the
memory M.sub.0, four points of the memory M.sub.1, two points of the
memory M.sub.2 and only one point of the memory M.sub.3. It should also be
noted that it is the same signal K as in the circuit of FIG. 2 that gives
the compression factors to the filter F of FIG. 3.
As in the case of FIG. 2, the filter F of FIG. 3 is followed by a memory Ms
unless the work is done in real time.
The present invention is not restricted to the example described. Thus, in
particular, the number of auxiliary filters, such as the filters F1, F2,
F3 of FIG. 3, may be different from three, it being understood that there
should be at least one auxiliary filter, and that, in this case, the
switching over by the change-over switch 2 will be done between two
memories, the image memory and the memory placed at the output of the
single auxiliary filter. Similarly, the compression rates of the auxiliary
filters may be other than 2, and may even be different from one another
for filters of one and the same device.
Top